LLM inference in C/C++
C++ ggml
llama : fix init in several tools/examples (#29632)
Georgi Gerganov6d78fb0727fd
diff --git a/examples/simple-chat/simple-chat.cpp b/examples/simple-chat/simple-chat.cppindex 97e9dc9842..30a0966e07 100644--- a/examples/simple-chat/simple-chat.cpp+++ b/examples/simple-chat/simple-chat.cpp@@ -65,8 +65,7 @@ int main(int argc, char ** argv) {}}, nullptr);- // load dynamic backends- ggml_backend_load_all();+ llama_backend_init();// initialize the modelllama_model_params model_params = llama_model_default_params();diff --git a/examples/simple/simple.cpp b/examples/simple/simple.cppindex 9f0a25d713..982a4d8600 100644--- a/examples/simple/simple.cpp+++ b/examples/simple/simple.cpp@@ -77,9 +77,7 @@ int main(int argc, char ** argv) {}}- // load dynamic backends-- ggml_backend_load_all();+ llama_backend_init();// initialize the modeldiff --git a/include/llama.h b/include/llama.hindex e808fa9ec1..dea4d46062 100644--- a/include/llama.h+++ b/include/llama.h@@ -489,6 +489,7 @@ extern "C" {LLAMA_API void llama_backend_free(void);// Optional: enable numa optimizations+ // TODO: deprecate and make part of llama_backend_init()LLAMA_API void llama_numa_init(enum ggml_numa_strategy numa);// Optional: an auto threadpool gets created in ggml if not passed explicitlydiff --git a/tests/test-fusion.cpp b/tests/test-fusion.cppindex 467248f0f1..65d444ecf4 100644--- a/tests/test-fusion.cpp+++ b/tests/test-fusion.cpp@@ -297,7 +297,7 @@ int main(int argc, char ** argv) {}common_init();- ggml_backend_load_all();+ llama_backend_init();ggml_backend_dev_t dev = ggml_backend_dev_by_name(device_name.c_str());if (!dev) {diff --git a/tools/cli/cli.cpp b/tools/cli/cli.cppindex dcdb6aeac2..92c86614f8 100644--- a/tools/cli/cli.cpp+++ b/tools/cli/cli.cpp@@ -41,6 +41,9 @@ int llama_cli(int argc, char ** argv) {return 1;}+ llama_backend_init();+ llama_numa_init(params.numa);+#if defined (__unix__) || (defined (__APPLE__) && defined (__MACH__))struct sigaction sigint_action;sigint_action.sa_handler = signal_handler;